Arcana: AI Supercharged Notetaking

Github and Collaboration

For a full overview, you can view the source code, add feature requests and more on our official github repo: Obsidian Arcana

We are actively looking for new collaborators so please hit me up on Obsidian Forum or Twitter@AFV_7


Arcana is a collection of AI powered tools designed to help you be more creative and productive with your Obsidian vault. Each tool is inspired by a famous historical figure:

  • Socrates - Conversation
  • Agatha Christie - Text Generation
  • Richard Feynman - Flashcard Generation
  • Charles Darwin - Auto Tagging
  • Nostradamus - Note Naming

Custom Agents

A USP of Arcana is its ability to support user defined custom agents. Consider an agent who is particulary good at teaching you based on preferred learning style and framework. Users can specify templates for conversation agents that they can then chat with.

Future Work

  1. Tool’s Tools - Empowering the AI agents with search and browsing tools that it can make use of to complete its tasks for you.
  2. UI Improvements
  3. A Youtube Overview
1 Like

Not sure that having a robot do my studying and research, instead of me, is an improvement.



I agree that it’s not an instant improvement in that using it like a shortcut won’t actually help you learn long-term.

However, this is still an amazing tool that, if used right, can make learning a lot more efficient!
Like, I have to study US Government right now, and there are so many terms to memorize. When a term comes up in my lesson notes, I make a link to a note with the term’s definition so that I won’t forget it later. It would be sooo nice to be able to make the term note and have the definition just generate itself.

Of course, you’d probably want to double check everything because, well, its AI. But when you have, like, 300 terms, this could definitely help you learn more efficiently.

And then flashcard generation! I spend soooo much time making flashcards when I ought to be studying them. This plugin would speed up the process.
And of course, the AI could miss a few things, and I’d have to fill in the gaps, but still—this would speed up my learning.

I just wish I had the tokens to try it out.

1 Like

That’s the problem, isn’t it? These robots make up answers that look correct but are lies, and then you end up not trusting anything because, well, if you knew at a glance the answers were wrong then you wouldn’t need the robot, would you? So either the user doesn’t care, or cares and spends a lot of time fact checking. Maybe this is just another Theranos, where some grifters make a lot of money hyping “world changing technology” that isn’t.

1 Like

That is true. Earlier I asked ChatGPT to format some Fountain text into Markdown. It did so incorrectly, so I tried to test if it actually knew what I was asking of it. Turns out, it knows what fountain is, but it does not know fountain. I gave it simple syntax to interpret, and it made up a bunch of stuff! Like, one was pretty creative. It was guessing. A lot of its guessing did turn out correct, but it was still guessing.

So yeah, it sort of does “lie” in a way. But I wouldn’t call it lying because, well, it’s a computer. It isn’t smart, and it’s knowledge is limited to what we humans feed it.

But, I think we should look at it like any source. When I come across a new term in my government class, first thing I do is insert the first section of the Wikipedia article on that term into the term note. Now, is Wikipedia 100% reliable? No. But it’s an accessible resource. So, I look at the article and ask myself, does this make sense in the context of the lesson? Often it’s yes. So it’s likely true. And that just saved me a bit of time, and will help me remember the term.

If it doesn’t make sense or seems kinda iffy, I look into other resources to find out what the real definition is. ChatGPT is kind of like that. No source is reliable, and you can’t really trust anything on the internet. Books can be incorrect. Articles from the world’s top scientists can be incorrect. That doesn’t mean you can’t use them, though. Just be aware that they can be incorrect, and look out for cues that they could be incorrect. And, when in doubt, look at other sources.

I still think that this is a great learning resource to take advantage of.

1 Like

I agree with others that AI text generation misses the point of notetaking

But that auto tagging feature is exciting! I’ve never seen a plugin that does that.

1 Like

I am of the philosophy of different strokes for different folks. For me, notetaking is about discovering and creating new ideas, and the part of transcribing existing material into one’s repository might be best outsourced to the machine.

An illuminating example I had the other day was to ask Christie to write ‘A list of grilling times for different fishes and meats’ that I stored in a ‘Grilling Times’ note. I consider this time saved as opposed to an opportunity lost.

With these tools, I hope that they become less of a hammer to make all else look like nails, and instead are incantations that one can summon selectively and in the best situation.

1 Like